How to stay at the page with AI, without compromising your sustainability values ?

AI can speed up design, engineering and project management work considerably, but it also carries a real energy and water cost that scales with how models are trained and used. The practical answer for a sustainability-minded business is not to avoid AI altogether, but to use it selectively, measure its footprint honestly, and offset it through the efficiency gains it creates elsewhere in the project. At SwissTechNova, we treat AI the same way we treat any new material or technology in a build: adopt it where it genuinely adds value, and be transparent about the trade-offs.

Where AI actually fits into an architecture and engineering process

In practice, AI tools are most useful for narrow, well-defined tasks rather than as a blanket replacement for expert judgment. In our own workflow across architecture and engineering projects, AI-assisted tools help with early-stage feasibility checks, drafting energy simulations, generating design variants for client review, and flagging risks in large datasets during project management. The human decisions — structural safety, material choice, regulatory compliance, client intent — remain firmly with our architects and engineers, who bring an average of 19 years of professional experience to every brief.

The key is to be deliberate: use AI where it removes repetitive or computational burden, and keep it out of decisions where craftsmanship, code compliance, or long-term durability are at stake.

How to measure the environmental impact of your AI use

Measuring AI's footprint is harder than measuring a building's energy performance, but a few practical proxies help:

• Compute intensity — how large and complex the model is, and how often you query it. Simple, task-specific tools generally use far less energy than large general-purpose models run repeatedly for the same task.
• Data centre energy mix — providers increasingly disclose whether their infrastructure runs on renewable electricity; choosing vendors that report this transparently is a reasonable proxy for impact.
• Frequency and necessity — many teams run the same prompt or simulation multiple times "just to check." Tracking usage volume against actual project value is often the simplest way to spot waste.

For a firm managing physical construction, this mirrors how we already track embodied carbon and energy performance in buildings — the same discipline of measuring before optimising applies to digital tools. The International Energy Agency has published useful benchmarks on the growing electricity demand from AI and data centres, which is a good external reference point when setting internal targets.

Setting a realistic target

There is no universal number that defines "sustainable AI use" for a design or engineering practice — the right target depends on your size, the tools you use, and how central AI is to your deliverables. A workable approach is to set a directional goal rather than a fixed figure: reduce redundant AI queries, prefer smaller or specialised models over large general ones for routine tasks, and reserve the most compute-intensive tools for the projects where they clearly outperform manual methods. Reviewing this target annually, alongside your other sustainability commitments, keeps it honest and adjustable as the technology and its energy data evolve.

Alternatives to heavy AI use — and how much difference they make

Not every task needs AI, and the alternatives often cost less in both money and energy:

• Established calculation software and BIM tools for structural and energy modelling, which are far less compute-intensive than generative AI for the same output.
• Standardised checklists and templates for feasibility studies, reducing the need to re-generate similar content repeatedly.
• Experienced human review for design and risk decisions, which remains more reliable than AI for the nuanced, code-specific judgment calls involved in complex engineering projects.

The difference these choices make is cumulative rather than dramatic on any single task — but across a full project lifecycle, avoiding unnecessary AI runs and defaulting to lighter tools adds up, much like specifying efficient materials adds up over the life of a building.

Possible scenarios for the future

Three broad scenarios seem plausible for how AI and sustainability will intersect in design and engineering practices over the next few years:

1. Efficiency gains outpace demand growth — as AI models become more energy-efficient per query, and data centres shift further toward renewable power, the environmental cost per use could fall even as adoption rises.
2. Demand grows faster than efficiency — if AI use in design, simulation and project management expands quickly, overall energy demand could still rise despite per-query improvements, making disciplined, selective use more important, not less.
3. Regulation and reporting standards mature — carbon reporting requirements may eventually extend to digital tools and services, similar to how construction already tracks embodied carbon; firms that already measure and manage AI use will be better positioned.

Given this uncertainty, the safest strategic position for a practice like ours is to keep AI use purposeful, documented, and reviewed — so we can adapt quickly whichever scenario plays out.